Impact of the Prevalence of Cognitive Impairment on the Accuracy of the Montreal Cognitive Assessment
| Autor: | Landsheer, J.A., Leerstoel Heijden, Methodology and statistics for the behavioural and social sciences |
|---|---|
| Přispěvatelé: | Leerstoel Heijden, Methodology and statistics for the behavioural and social sciences |
| Rok vydání: | 2020 |
| Předmět: |
Optimal cutoff
prevalence Sample (statistics) Odds uniform data set 03 medical and health sciences 0302 clinical medicine Predictive Value of Tests classification accuracy Statistics Humans Medicine Cognitive Dysfunction 030212 general & internal medicine Cognitive impairment Aged cognitive impairment business.industry Montreal Cognitive Assessment Original Articles test accuracy Mental Status and Dementia Tests Test (assessment) body regions Data set Psychiatry and Mental health Clinical Psychology Predictive value of tests National Alzheimer’s Coordinating Center ComputingMethodologies_DOCUMENTANDTEXTPROCESSING Alzheimer disease Geriatrics and Gerontology business Gerontology 030217 neurology & neurosurgery MontrealCognitive Assessment |
| Zdroj: | Alzheimer Disease and Associated Disorders Alzheimer disease and associated disorders, 34(3), 248. Lippincott Williams and Wilkins Ltd. |
| ISSN: | 0893-0341 1546-4156 |
| DOI: | 10.1097/wad.0000000000000365 |
| Popis: | Supplemental Digital Content is available in the text. Objectives: The focus of this study is the classification accuracy of the Montreal Cognitive Assessment (MoCA) for the detection of cognitive impairment (CI). Classification accuracy can be low when the prevalence of CI is either high or low in a clinical sample. A more robust result can be expected when avoiding the range of test scores within which most classification errors are expected, with adequate predictive values for more clinical settings. Methods: The classification methods have been applied to the MoCA data of 5019 patients in the Uniform Data Set of the University of Washington’s National Alzheimer’s Coordinating Center, to which 30 Alzheimer Disease Centers (ADCs) contributed. Results: The ADCs show sample prevalence of CI varying from 0.22 to 0.87. Applying an optimal cutoff score of 23, the MoCA showed for only 3 of 30 ADCs both a positive predictive value (PPV) and a negative predictive value (NPV) ≥0.8, and in 18 cases, a PPV ≥0.8 and for 13 an NPV ≥0.8. Overall, the test scores between 22 and 25 have low odds of true against false decisions of 1.14 and contains 55.3% of all errors when applying the optimal dichotomous cut-point. Excluding the range 22 to 25 offers higher classification accuracies for the samples of the individual ADCs. Sixteen of 30 ADCs showed both NPV and PPV ≥0.8, 25 show a PPV ≥0.8, and 21 show an NPV ≥0.8. Conclusion: In comparison to a dichotomous threshold, considering the most error-prone test scores as uncertain enables a classification that offers adequate classification accuracies in a larger number of clinical settings. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|